In modern power grids, markets are employed to match intermittent generation, traditional dispatchable generation, and loads that assume reliable power. For example, in MISO, with a collection of off-line, day-ahead and real-time market structures, power is sold and allocated in 5-minute intervals. Despite sophisticated participants and markets, these systems suffer from significant curtailment and uneconomic generation – stranded power.

We characterize the duty facter, geographic distribution, and quantity of stranded power. Using recent MISO power grid data (28 months, 77M pricing intervals), we show that stranded power is vast (7.7 TWh in 2014), can be made useful. We demonstrate acquisition models that achieve 20-80% duty factors. Simulations of future grid configurations suggest stranded power will increase dramatically as RPS increases (50% in California and New York for 2030, and DOE’s Wind Power vision for a 35% national RPS for 2050). A wide range of challenging questions about the robustness and scaling of stranded power remain open.

We propose the use of large-scale computing, Zero-carbon Cloud, as a “dispatchable load” to enhance both grid stability and efficiency. These computing structures have been shown to be productive and cost-effective when supplied with stranded power, and are thus a promising new approach really reduce the carbon footprint of cloud computing. With advanced power grid simulation techniques, we also study the impact of both traditional cloud data centers and ZCCloud’s new “dispatchable” loads on grid stability and efficiency. A wide range of challenging questions about how zero-carbon cloud can deliver a wide-array of valuable cloud services remain open.

More information is available at